Zero‐Altered and other Regression Models for Count Data with Added Zeros

Abstract

On occasion, generalized linear models for counts based on Poisson or overdispersed count distributions may encounter lack of fit due to disproportionately large frequencies of zeros. Three alternative types of regression models that utilize all the information and explicitly account for excess zeros are examined and given general formulations. A simple mechanism for added zeros is assumed that directly motivates one type of model, here called the addcd-zero type. particular forms of which have been proposed independently by D. LAMBERT (1992) and in unpublished work by the author. An original regression formulation (the zero-altered model) is presented as a rcduad form of the two-part model for count data, which is also discussed. It is suggested that two-part models be used to aid in development of an added-zero model when the latter is thought to be appropriate.

Nectar bacteria, but not yeast, weaken a plant–pollinator mutualism

“…Because of a zero-inflated distribution [28], a two-part approach [29,30] was taken to analysing seed set data. First, we used a logistic regression to model the probability of setting seeds (yes or no), with the same predictors as for stigma closure, followed by a LR test.…”

Inter-annual variability in Prosopis caldenia pod production in the Argentinean semiarid pampas: A modelling approach

“…The explanatory covariates selection was carried out by first independently fitting a binomial regression model for the dichotomized data for fruiting occurrence and, after, fitting a log-normal model using only the non-null intensity data (weight of pods), as proposed by Heilbron (1994) and Woollons (1998). The independent fitting of these generalized linear models can be accomplished using maximum likelihood estimation methods.…”

Quantitative Analysis of Complex Tropical Forest Stands: A Review

“…The works of White and Bennetts (1996) give an example of fitting the negative binomial distribution to point count data for orange-crowned warblers (Vermivora celata) when comparing their relative abundance among forest sites. Zero-inflated Poisson (ZIP) models and negative binomial regression models are recommended for analysis of count data with frequent 0 values (e.g., rare species studies) in which data transformations are not feasible ( Heilbron 1994, Welsh et al 1996, Ridout et al 1998, Agarwal et al 2002, Hall and Berenhaut 2002.…”